1. Field of the Invention
The present invention relates to a substrate material for use as, e.g., a heat sink material in a semiconductor device, and also to a substrate, a semiconductor device, and method of producing the same.
2. Description of the Prior Art
With the recent remarkable increases of the processing rate of semiconductor devices and the degree of integration in semiconductor devices, the heat generated by semiconductor elements has come to produce influences that are not negligible. As a result, substrate materials for mounting semiconductor devices have come to be required to have a high thermal conductivity for efficiently removing the heat generated by semiconductor elements.
Substrate materials are further required not to be deformed by a thermal stress at the interface between the substrate materials and other device members used in combination therewith. Hence, substrate materials are required not to have a large difference in thermal expansion coefficient with semiconductor chips, packages, etc. used in combination therewith. Specifically, since the thermal expansion coefficients of silicon and GaAs, both used as semiconductor elements, are as low as 4.2.times.10.sup.-6 /.degree. C. and 6.5.times.10.sup.-6 /.degree. C., respectively, and that of Al.sub.2 O.sub.3, which is widely used as a packaging material, is also as low as 6.5.times.10.sup.-6 /.degree. C., substrate materials desirably have low thermal expansion coefficients which are almost the same as the above values.
In recent years, plastics have come to be increasingly used as packaging materials in place of ceramics such as Al.sub.2 O.sub.3. In the case of using a plastic package, the substrate material used therewith can have a higher thermal expansion coefficient than in conventional semiconductor devices, because the plastic has a high thermal expansion coefficient and a semiconductor element is bonded thereto with a resin. Namely, the optimal range of the thermal expansion coefficients which semiconductor substrate materials are required to have is from 7.times.10.sup.-6 to 20.times.10.sup.-6 /.degree. C., although it varies depending on combinations with other device members including packaging materials; those values are high and that range is wide as compared with the case of ceramic packages.
Conventionally in the case where a low thermal expansion coefficient is necessary, substrate materials for carrying a semiconductor device of a Cu--W composite alloy or a Cu--Mo composite alloy have been frequently used. Since these alloys can be regulated in thermal expansion so as to be suitable for use with plastics by controlling the amount of copper or molybdenum, they can be also applicable to with plastic packages. However, because of their low rigidity, the plastics are apt to deform when used in combination with materials having a high specific gravity, such as Cu--W alloys and Cu--Mo alloys, and this limits the use of these alloys as substrate materials in combination with plastic packages.
For electrically connecting a semiconductor element to a package, a technique of using solder bumps in place of wires (flip chip bonding) and a technique of using solder balls in place of pins for connection to a base substrate (ball grid array bonding) have come to be widely used. These techniques also have difficulties in application to substrate materials made of a Cu--W alloy or a Cu--Mo alloy, because the use of such heavy substrate materials may flatten the solder balls excessively. In addition, the use of substrate materials made of these alloys is disadvantageous in cost because tungsten and molybdenum are relatively expensive metals.
On the other hand, in the case where a high thermal expansion coefficient is desired, substrate materials made of Al or Cu, which are inexpensive metals, or of an alloy of these have been frequently employed. However, these semiconductor substrate materials have the same problem as the Cu--W alloys or the like because Cu also has a density as high as 8.9 g/cm.sup.3. Substrate materials made of aluminum are free from the above problem when used in combination with plastic packages and connected by ball grid array bonding, since the density of aluminum is as low as 2.7 g/cm.sup.3. However, substrate materials made of aluminum have problems that they can be used in only limited applications because aluminum has a thermal expansion coefficient as high as 23.5.times.10.sup.-6 /.degree. C., and that the substrates are apt to warp or deform because of their low rigidity.
Under these circumstances, there is a desire for a substrate material which not only can be regulated so as to have any thermal expansion coefficient in the wide range of from 7.times.10.sup.-6 to 20.times.10.sup.-6 /.degree. C., especially from 7.times.10.sup.-6 to 15.times.10.sup.-6 /.degree. C., but also has high heat dissipation properties and is lightweight. It is thought that substrate materials should generally have a thermal conductivity of at least 100 W/m.multidot.K. However, there are increasing cases where the semiconductor substrate materials used in combination with plastic packages, having a poor thermal conductivity, are required to have a thermal conductivity of 180 W/m.multidot.K or higher when heat dissipation from the whole package is taken in account.
Aluminum composite alloys were recently proposed as materials which are lightweight and satisfy the above-described requirements concerning thermal expansion coefficient and thermal conductivity. Among these, use of an aluminum/silicon carbide (Al--SiC) composite alloy as a substrate material is being investigated because the starting materials, i.e., aluminum and silicon carbide, both are relatively inexpensive and highly thermally conductive, and because a wide range of thermal expansion coefficients can be obtained by combining silicon carbide, having a low thermal expansion coefficient of 4.2.times.10.sup.-6 /.degree. C., with aluminum, having a high thermal expansion coefficient of 23.5.times.10.sup.-6 /.degree. C., in various proportions.
Conventionally employed processes for producing such an aluminum/silicon carbide composite alloy include the casting method disclosed, e.g., in Tokuhyo-Hei-1-501489 (unexamined published PCT application), the impregnation method described, e.g., in JP-A-2-243729 (the term"JP-A" as used herein means an "unexamined published Japanese patent application"), and the pressure casting method disclosed, e.g., in JP-A-61-222668.
For use in fields where high reliability is required, the substrate materials obtained are generally subjected to a surface treatment before being used as semiconductor substrates.
In order to use an aluminum/silicon carbide composite alloy as a substrate material for mounting a semiconductor device, the above-described production methods each has problems which should be solved.
First, the casting method has a drawback that the deviation of composition which is caused during cooling is difficult to avoid. This is because since the Al--SiC alloy produced by the casting method necessarily has a higher aluminum concentration in the surface part, the difference in silicon carbide concentration between the central and surface parts exceeds 1% by weight, making it impossible to obtain a material having a homogeneous composition. It is also difficult to completely eliminate voids. Although the pressure casting method is effective in eliminating most voids, the concentration of aluminum around the surface tends to be high due to the pressure applied. It is hence difficult in the pressure casting method to reduce the difference in silicon carbide concentration between the central and surface parts to 1% by weight or smaller.
On the other hand, the impregnation method in which a preform of silicon carbide is impregnated with molten aluminum has a drawback that aluminum should be infiltrated in an excess amount in order to obtain a completely dense alloy. Consequently, the alloy obtained has the excess aluminum on the periphery thereof and cannot have the original shape of the preform before impregnation. It is therefore difficult to obtain a substrate material having satisfactory dimensional precision. For obtaining the desired dimensions, it is necessary to conduct an operation for removing the excess aluminum from the whole periphery. In the pressure infiltration method, in which a preform of silicon carbide is placed in a pressure vessel and molten aluminum is forced into the vessel, the resulting alloy has an aluminum film corresponding to the clearance between the preform and the pressure vessel. Since this aluminum film is uneven in thickness, it not only impairs the low thermal expansion coefficient of the alloy material, but is causative of warpage, etc.
The casting method has another drawback that since the method involves the step of casting a molten metal, the concentration of aluminum in the melt should be at least 70% by weight. In the pressure casting method also, a melt having a silicon carbide concentration not lower than 30% by weight has poor flowability and has been unsuitable for use in producing alloys of complicated shapes, resulting in poor production efficiency. It has been found that alloys which can be produced by the impregnation method so as to retain the same dimensions as the preform and to be optionally dense have compositions in which the concentration of silicon carbide is around 70% by weight. This is because if an alloy having a silicon carbide concentration lower than 70% by weight is to be produced, the silicon carbide preform has a reduced strength and is hence apt to deform or warp during impregnation with aluminum or during subsequent cooling due to a difference in thermal expansion coefficient. As a result, the alloy obtained hardly retains the same dimensions as the preform.
Further drawbacks of the impregnation method are that since the framework of the alloy produced is constituted of silicon carbide, warpage correction by sizing after alloying is impossible, and that the processing of the alloy is possible only by grinding with a diamond wheel and is hence costly. The pressure casting method has a further problem that it has a far higher equipment cost than the casting and impregnation methods. Therefore, it has been difficult to produce an aluminum/silicon carbide composite alloy having a homogeneous composition in which the silicon carbide concentration is higher than about 30% by weight but not higher than about 70% by weight at low cost by any of those prior art methods.
Substrate materials for use in plastic packages of semiconductor devices are frequently required to have a thermal expansion coefficient of about from 7.times.10.sup.-6 to 13.times.10.sup.-6 /.degree. C. for the reasons as set forth above. In order for an aluminum/silicon carbide composite alloy to meet this requirement, the alloy should have a silicon carbide concentration of from 50 to 70% by weight. It has however been difficult to produce an alloy having such a composition at low cost by any of the casting, impregnation, and pressure casting methods described above. It has also been difficult to obtain an aluminum/silicon carbide composite alloy having a thermal conductivity of 180 W/m.multidot.K or higher by any of those prior art methods, except in the case of alloys having silicon carbide concentrations exceeding 70% by weight or below 10% by weight.
Although an aluminum/silicon carbide composite alloy as it is can be used as a semiconductor substrate material, it is desirably subjected to a surface treatment when the composite alloy is to be used in fields where high reliability is required, e.g., in work stations. For use in these fields, semiconductor substrate materials are required not to suffer any change in properties, e.g., thermal conductivity, or in appearance, etc., through various reliability tests such as a thermal cycling test in which the substrate materials are repeatedly exposed to -65.degree. C. and 150.degree. C., a PCT (pressure cooker test) in which the substrate materials are exposed to an atmosphere of 121.degree. C., 100% RH, and 2 atm., and an HAST (highly accelerated stress test) in which the substrate materials are exposed to an atmosphere of 125.degree. C., 85% RH, and 2 atm. However, the exposed aluminum part on the surface discolors through these tests. Discoloration is apt to occur also at the Al/SiC interface. Hence, the substrate materials undergo a considerable change in appearance through the above tests. It is therefore necessary to perform a surface treatment suitable for aluminum/silicon carbide composite alloys and to use a technique for the treatment.
The surface state of a substrate material for mounting a semiconductor device is important, because it influences the flow of solder when a semiconductor chip or a package is fixed to the substrate formed by the material by using solders. For use with some solders, a surface treatment suitable therefore is necessary in order to ensure reliability with respect to the strength and other properties of soldering parts. The surface state is important also in bonding with a resin, which technique is becoming the mainstream recently. Since different surface states result in different strengths of resin bonding, a surface treatment suitable for the desired bonding strength is required.
Aluminum/silicon carbide composite alloys have a high degree of hardness. Therefore it is very difficult to form shape, especially a complex shape such as heatsink by using aluminum/silicon carbide composite alloys. It is required that a substrate material is able to be formed easily to be near net shape precisely.